Image to PromptImage to Prompt
FREERuns on your deviceRuns on your deviceSee exactly what a model would see in your photo, then copy it as a prompt.Free vs Pro Same Tool, More Power.
| Feature | Free | Pro |
|---|---|---|
| Prompts per photo | Unlimited | Unlimited |
| Swap in my product | — | ✓ Keeps the scene, changes the subject |
| Send to Creatide | ✓ | ✓ |
From Photo to Prompt in Three Steps.
- Add a photoDrop a JPG, PNG or WebP. It’s read by a small model running in your browser and never uploaded.
- Pick a modelChoose where you’ll paste it. The prompt is reordered and trimmed for that model.
- Copy or createCopy the prompt, or send it straight to Creatide to make the image or video.
What the Model Actually Reads.
Every prompt we write is built from the same six parts. Each one is colored the same in the tool, so you can see which words came from which part of your photo.
Subject
The main thing in the frame, described by material, color and shape. “A green glass bottle” works better than “a product”.
Setting
The surface and the space around the subject. Models use this to place shadows, reflections and scale.
Light
Direction, softness and color of the light. This single line changes the mood of a shot more than anything else.
Camera
Angle, distance and lens. Eye-level 50mm reads as honest and natural; a low wide angle makes the subject feel bigger.
Style
Editorial, film, 3D, illustration. Name one look, not five, or the model averages them into something flat.
Colors
The main colors as names or hex codes, so a brand palette survives into the new image.
One Photo, Written for Each Model.
Models read prompts differently. Pick a model in the tool and the same facts are reordered and trimmed to fit it.
| Model | Best order | Length | Tip |
|---|---|---|---|
| Creatide, Meta Muse Image, Grok Imagine 2.0, Creatide AI Muse, Grok Imagine, Runway, Z Image | Subject → setting → light → camera → style | Full sentences | Add the product name and your brand colors |
| Midjourney | Subject and style first, details after | Short phrases | Fewer words, stronger style words |
| Qwen Image 3, Qwen Image 2.1, Wan 2.7, GPT Image 2.6, Seedream v5.0, FluX 3 Klein, FluX 3, Kling Image O3, Nano Banana 2.1, Ideogram 4.5 | Natural description, front to back | Full sentences | Describe text on signs or labels in quotes |
The table lists the image models live on Creatide and follows them as the list changes. A model outside these families gets the Creatide prompt.
Related Tools
18 tools · scroll
Video01Video CompressorShrink a video to fit Discord, WhatsApp or an email.
Video02Trim, Cut & MergeCut the slow parts, join clips, rotate or flip.
Video03Slow Motion & Speed RampSlow down the best moment, trim the rest, smooth the ramps.
Video04Extract Audio or MuteSave a video's sound as MP3, or the video without it.
Video05Frame GrabberSave the sharpest frame of a video as a still image.
Subtitles06Video to TextGet a transcript of any video or audio, with timestamps.
Music & audio07Text to SpeechTurn a script into a natural voiceover and save it as MP3.
Music & audio08Slowed + ReverbMake slowed, reverb, sped up and nightcore versions of a song.
Image09Remove Image BackgroundCut out a product or a person, ready for any design.
Image10Image UpscalerMake small or blurry photos sharp, for free.
Image11Magic EraserBrush over people, objects or text and they’re gone.
Video12Video to PromptA prompt for every scene of a video, ready to remake.
Video13AI Clip StitcherJoin AI clips with matched color and smooth cuts.
Video14Seamless Loop MakerMake a clip loop forever without a visible jump.
Video15Video Background RemoverCut a person out of a video, with a transparent background.
Video16Color MatchGive every clip the same look, or save it as a LUT.
Music & audio17Voice CleanerTake the noise and hum out of a voice recording.
Image18Start Frame PrepFit a start frame to the exact size an AI video model wants.Image to Prompt, Built for AI Image and Video
Image to Prompt looks at a photo and writes down what an AI model needs to recreate it: subject, setting, light, camera and style. Each part is colored so you can see where every word came from and edit only the part you want to change.
Use it to copy the look of a reference, to describe your own product consistently across many images, or to write the first frame of a video before you animate it in Creatide.
Your Photo Stays on Your Device
The description is made by a compact vision model that runs inside your browser with WebGPU. The first run downloads the model once; after that it works offline and nothing is uploaded.
On older devices without WebGPU, the tool uses the slower WASM mode instead of sending your photo anywhere.
Turn a Reference into Your Own Shot.
The fastest way to a good AI image is to copy what already works and change one thing.
- 01
Start from a Reference You Like
Use a photo whose light and framing you want to repeat, even if the product in it isn’t yours.
- 02
Keep the Look, Change the Subject
Replace only the subject line with your own product. Light, camera and style stay, so the new image matches the reference.
- 03
Lock Your Colors
Turn on hex colors so the palette carries over exactly, then change one color at a time if you want a variation.
- 04
Animate It
Add a short motion line, like “slow push in, steam rising”, and paste the prompt into Creatide’s video generator with the photo as the first frame.
Image to Prompt, Answered.
How does Image to Prompt work?
A small vision model reads the photo in your browser and lists the subject, setting, light, camera angle, style and main colors. We then put those parts in the order the model you picked reads best.
Is my image uploaded?
No. The model runs on your device. The first time, your browser downloads the model files; after that, nothing leaves your device.
Which AI models is the prompt written for?
Every image model live on Creatide, on the free plan too. Midjourney, FLUX, GPT Image, Qwen and Seedream each get the order they read best; any other model gets the Creatide prompt. The list follows the models on Creatide.
Will the AI make the exact same image?
No prompt reproduces a photo exactly. You’ll get the same kind of shot: similar subject, light, angle and colors. For a closer match, use the photo itself as a reference image in Creatide.
Can I use someone else’s photo as a reference?
You can describe a photo’s look to learn from it, but don’t present someone else’s work as your own, and only upload images you have the right to use.
Why are the colors written as hex codes?
Hex codes keep brand colors exact. Color names like “warm beige” leave the model room to drift. You can switch hex codes off.
Does it work for people?
It describes clothing, pose, setting and light, but not who someone is. It never names people in photos.
What does Swap in my product do?
Pro keeps every part of the prompt except the subject and replaces it with your product name and description, so you can reuse a reference look for your own items.
Does it work on a phone?
Yes, on recent iPhones and Android phones with a modern browser. The first download is about 215 MB, so use Wi-Fi.
Start generatingThe Next Shot IsJust a Prompt Away
Unlock premium AI creation, faster rendering, commercial rights and watermark-free exports whenever your next idea is ready.